Background on the NSF/CEME Decentralization Conference Series

The following description of the Decentralization Conference Series is based on my impressions as well as some very helpful instruction from Ted Groves, David Kreps, and Tom Marschak. In writing the description below, I have mixed my impressions with parts pillaged from some of their emails. If you have anything to add to this (or correct in it), please feel free to email me at
Thank you, Matt Jackson

The NSF/CEME Decentralization Conference Series is funded by a grant from the National Science Foundation that is administered through the NBER. CEME stands for Conferences on Econometrics and Mathematical Economics. The purpose of CEME was to create several series of standing national seminars on various topics. Ken Arrow was the original PI on the mathematical economics side, and on that side two series were begun: one was the General Equilibrium conference series (coordinated by Gerard Debreu), and the other was the Decentralization conference series (first coordinated by Roy Radner). Their successors were David Kreps as PI, Don Brown as the GE series coordinator and Ted Groves as the Decentralization series coordinator; and more recently Jose Scheinkman as PI, Chris Shannon as the GE series coordinator and me as Decentralization series coordinator. These conferences are traditionally run once a year, and occasionally more often. The first Decentralization Conference was held at UC Berkeley in 1970 (or perhaps 1971 - I will do some digging).

Much of the Decentralization Conference Series has focussed on questions of design of various systems for production and allocation of goods, services, and information. These range from understanding voting systems to markets to the organization of the internal lines of communication and the assignment of tasks within a firm. A partial list of topics covered at past conferences includes: mechanism and market design, resource allocation, matching, team theory, contracting, organization of the firm, communication, computation, and processing of information.

Roy Radner was likely the source of the name for the conference series, in addition to being its first coordinator. My impression is that the word decentralization reflects the fact that the starting point in many of the problems addressed by the series is that the necessary information starts in a decentralized state - for example: peoples' preferences in a voting setting, who has what goods and wants to buy what in a market, and who knows what about demand, costs, etc., in a firm. In some cases, the term decentralization is partly a misnomer, since some of the mechanisms or institutions used to solve these problems (e.g., an auction) bring the critical actors together in some way or at least some of the calculations are centralized; while other applications really are completely decentralized. But I think that perhaps the history of the conference reflects the fact that these systems were viewed as alternatives to centralized or planned economies when the conference series was first funded, during the cold war. Those boundaries and distinctions have faded over the years, but the name has remained.

A first explicit and careful definition of the term ``decentralization'' with respect to processes for aggregating information for the allocation of resources appears in Tom Marschak's dissertation (1959 - Econometrica) ``Centralization and Decentralization in Economic Organization.'' This, along with the Leo Hurwicz article (1994 - Economic Design) ``Economic Design, Adjustment processes, mechanisms, and institutions,'' provide a nice view of how this decentralization literature grew out of the Barone, Mises, von Hayek, Lange and Lerner debates over the feasibility of a centralized socialist economy. The roots of the question of how to collect decentralized information and allocate resources can be traced to the early Walrasian tatonnement processes, and the later Tjalling Koopmans' (1951) formalization of adjustment processes as well as Arrow-Hurwicz gradient process. Early uses of the term ``decentralization'' can also be found in Hurwicz's '59 paper, "Optimality and Informational Efficiency in Resource Allocation Processes", as well as the late fifties work of Arrow and Hurwicz (see the Marschak paper for references). These set the stage for the conference series with a set of important issues.

Much of the early work on decentralization concerned the convergence properties and complexity of the systems involved. The Marschak Festschrift "Decisions and Organization", edited by Roy Radner and Bart McGuire (another early participant in the conference series) dates from the founding days of the conference, and gives a good view of some of this early work on decentralization. One of the classic references eventually coming out of this literature is Ken Mount and Stan Reiter's (1974-JET) paper ``The informational size of message spaces.'' While the work continued (and continues) on purely informational questions, the issue of incentives began to creep in as well, as it became clear that any system for making decisions or allocation resources might be open to some manipulation by its participants. Early mention of incentive issues, and perhaps the first coining of the term ``incentive compatibility'' come from Hurwicz's 1959 article. The fuller treatment of incentives then came into its own in the classic papers of Hurwicz's (1972) ``Decision and Organization,'' (which circulated in various incarnations before the conference series started) and Groves (1970) dissertation (which also circulated as a working paper before the conference series got underway, and was later partly published as the (1973-Econometrica) article ``Incentives in Teams,''), and eventually grew into the more modern theories of mechanism design and implementation. In fact, at the first decentralization conference Radner made a point of inviting two papers addressing incentives, presented by Groves and Hurwicz, and so incentive issues were on the table from the beginning of the conference series. Early discussion of information and incentives also came from the public goods literature, e.g. Samuelson (1954 - Review of Economics and Statistics) and Clarke (1971 - Public Choice), as well as the early auctions work of Vickrey (1961 - Journal of Finance). These eventually came under the umbrella of the mechanism design literature that stemmed from the decentralization series. In its current incarnation, the decentralization series now covers pretty much the design of any institutions, contracts, organizations, or relationships in which information and/or incentives play a role.

To return to the decentralization conference web page go to on.html
To return to the Matthew Jackson's home page go to
Last Modified: July 20, 2006